WO2005088464A2 - Circuit de traitement de donnees dans lequel des unites de traitement de donnees communiquent via un reseau - Google Patents

Circuit de traitement de donnees dans lequel des unites de traitement de donnees communiquent via un reseau Download PDF

Info

Publication number
WO2005088464A2
WO2005088464A2 PCT/IB2005/050626 IB2005050626W WO2005088464A2 WO 2005088464 A2 WO2005088464 A2 WO 2005088464A2 IB 2005050626 W IB2005050626 W IB 2005050626W WO 2005088464 A2 WO2005088464 A2 WO 2005088464A2
Authority
WO
WIPO (PCT)
Prior art keywords
stream
message
data processing
messages
node
Prior art date
Application number
PCT/IB2005/050626
Other languages
English (en)
Other versions
WO2005088464A3 (fr
Inventor
Edwin Rijpkema
Original Assignee
Koninklijke Philips Electronics N.V.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Koninklijke Philips Electronics N.V. filed Critical Koninklijke Philips Electronics N.V.
Priority to KR1020067017642A priority Critical patent/KR101125639B1/ko
Priority to US10/598,299 priority patent/US7839774B2/en
Priority to EP05708805A priority patent/EP1726132A2/fr
Priority to CN2005800066830A priority patent/CN1926826B/zh
Priority to JP2007501388A priority patent/JP4598051B2/ja
Publication of WO2005088464A2 publication Critical patent/WO2005088464A2/fr
Publication of WO2005088464A3 publication Critical patent/WO2005088464A3/fr

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00—Traffic control in data switching networks
    • H04L47/10—Flow control; Congestion control
    • H04L47/24—Traffic characterised by specific attributes, e.g. priority or QoS
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00—Data switching networks
    • H04L12/28—Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00—Traffic control in data switching networks
    • H04L47/10—Flow control; Congestion control

Definitions

  • Data processing circuit wherein data processing units communicate via a network.
  • the invention relates to a data processing circuit wherein data processing units communicate via a network, and in particular to an integrated circuit containing such a data processing circuit.
  • a signal processing circuit processes streams of real time, or near real time data such as audio or video data.
  • streams of such data are exchanged between different data processing circuits such as programmable processor circuits, dedicated processing circuits, peripheral IO circuits etc that process the streams.
  • the real time nature of the streams requires that a guaranteed communication bandwidth be available for the streams, along which a predetermined quantity of data can be transmitted per time unit.
  • the simplest way to provide a communication structure for such a purpose is the implementation of dedicated connection conductors for each stream. However, in most cases it is desirable to provide a communication structure that can be shared by different streams.
  • a communication bus structure suffices to provide for communication in simple circuits with a few data processing units.
  • communicating data processing units are coupled to a shared set of communication conductors, and data from different streams is passed over these communication conductors in alternate time-slots.
  • communication networks are used increasingly to interconnect data processing units inside integrated circuits. In a network there is no longer a single shared set of communication conductors that connect data processing units directly.
  • a network typically makes it possible to establish a connection between a pair of data processing units along a selectable one from several alternative paths. If part of one path is occupied, another path may be selected.
  • streams can be communicated between many arbitrarily selectable pairs of data processing units in parallel.
  • the real time nature of streams requires that a guaranteed amount of bandwidth, typically a predetermined fraction of available communication time-slots, be persistently available for each active stream.
  • Interruptions of stream communication due to contention for resources such as communication conductors should be prevented, because interruptions lead to malfunctioning, such as interruptions in video display, or hick-up in audio output etc.
  • the contention problem can be mitigated by temporary buffering of data in the nodes, but the amount of buffering should be minimized, because the inclusion of buffer memory increases the integrated circuit area used by the nodes.
  • Guaranteeing the persistent availability of real time bandwidth for a stream conventionally requires some form of centralized planning, or at least programming of the nodes in the network to reserve resources for each stream, e.g. to indicate that access to a communication conductor to a next node is reserved for a particular stream during predetermined time-slots.
  • it is a further object to support the use of data processing units that are programmed to execute a combined task in a distributed way, communicating data streams involved in the tasks over an interconnecting network and in which the programs are independent of the length of paths through the network.
  • the invention provides for a circuit according to Claim 1.
  • data processing units are interconnected through a time-slot multiplexing network that contains node circuits that forward messages through the network.
  • the data processing units select starting time slots of streams of messages that occupy resources in the network in a periodically repeating pattern of time-slots, each stream having the same period (called the network period), successive messages of a stream travelling along the same stream specific path through the network.
  • the node circuits arbitrate whether messages get access to sharable network resources (such as communication lines between node circuits) on the basis of a seniority of the messages in their particular streams (i.e. the number of messages that has preceded the message since the start of the particular stream).
  • a node circuit forwards a message from a particular stream only if messages with higher seniority from previously started streams leave sufficient resources.
  • the seniority may be indicated for example in a field in the messages, which is read by the node circuits to determine seniority. In this way, once an initial message has been forwarded from a node circuit, it is ensured that all subsequent messages will be forwarded from that node circuit, since newly arising periodic streams that contend for resources an integer number of network periods later will always be less senior.
  • a confirmation is sent back that will confirm arrival of the initial message of a stream at a destination data processing unit of the stream. Once the source data processing unit receives this confirmation one can be sure that all subsequent messages of the stream will be forwarded.
  • the decision to forward the messages can be simplified, by granting unconditional access if a message has a seniority beyond a threshold seniority of the last message before reception of the confirmation.
  • partial forwarding of the initial message of a stream by node circuits in the network is confirmed to the source data processing unit even before the initial message reaches the destination data processing unit. This reduces the threshold seniority after which no arbitration is needed.
  • node circuits send confirmations so frequently that the threshold seniority can be reduced to one, so that contention for resources has to be arbited only for the initial message of the stream. This simplifies arbitration and minimizes useless resource occupation.
  • the node circuits discard messages of a stream if they do not timely receive confirmation of forwarding of the initial message.
  • node circuits send back one or more confirmations to indicate that a message has been successfully forwarded through the network, back along the path of the messages of the stream.
  • the confirmations may be sent back outside the network for example.
  • the node circuits preferably send back the confirmations in return time slots that are selected so that crossing time slots, wherein the confirmation crosses subsequent messages of the stream, can be predicted by a predetermined scheme (e.g. so that crossing occurs in predetermined time slots, or in time slots that can be computed from information that is known to be included in the subsequent messages of the stream).
  • the node circuits use the predetermined scheme to predict crossing time slots, and test whether they have received crossing messages and confirmations for the same stream in the predicted crossing time slot. If an expected crossing of a message and a confirmation is not detected, the node circuit discards that message.
  • the confirmations are sent sufficiently frequently that the first crossing at a node circuit must occur at most one network period (and preferably exactly one network period) after that node circuit has forwarded the initial message of the stream.
  • the second and further messages of the stream can be forwarded without further arbitration if they cross with the confirmations.
  • a delay of at most one network period between forwarding at a node and subsequent crossing at that node can be ensured by sending successive confirmations of forwarding of the initial message in time-slots at most half a network period from each other.
  • the sum, of the time for the initial message to travel from a node circuit from which a first confirmation is sent to a further node circuit from which a second confirmation is sent and for the conformation to travel back from the further node circuit to the node circuit is at most a network period.
  • the confirmations are sent in return time slots selected so that successive confirmations occupy network each resource periodically with said network period. In this way, the seniority arbitration scheme can also be used for the confirmations.
  • Periodic occupation of network resources can be realized by sending confirmations of forwarding of an initial message when the initial message is forwarded in each of a series of time-slots that repeats with exactly half the network period.
  • the return time slots of sending back confirmations can be selected simply by counting clock periods.
  • a return stream is established from the destination data processing unit of an original stream in response to reception of the initial message of the original stream.
  • the destination data processing is arranged to send return messages of the return stream so that they follow the confirmations of the original stream after an integer number of network periods, starting from one network period after the last confirmation.
  • the preceding confirmations can be used as more senior messages of the return stream, which reduces (or even eliminates) the possibility that the return messages of the return stream will be discarded upon arbitration.
  • the return stream is guaranteed access without need to arbitrate forwarding of return messages.
  • the receiving data processing unit is pre-arranged to send the messages of the return stream periodically starting form a time-slot that is selected dependent on the time-slot of arrival of the initial message of the original stream. The selection is made so that a sum of the time-slot sequence numbers t a , t r of arrival of the initial message of the original stream and transmission of the message of the return stream equals a predetermined number "s" modulo the network period of the streams.
  • time-slots messages of the return stream and the original stream will cross each other at a node circuit.
  • the double of the sequence number of such crossing time- slots equals 2*(s+l), twice the predetermined number plus one modulo the network period of the streams.
  • the predetermined number may be selected the same for each stream, or it may be selected individually for each stream, the initial message of the stream indicating the number so that the node circuits can determine when to send confirmation messages.
  • the predetermined number equals minus one modulo the network period. In this way the node circuits can send confirmations in time slots whose sequence number equals zero modulo half the network period.
  • Figure 1 shows a data processing circuit
  • Figure 2 shows a path through a network, connecting a pair of data processing units
  • Figure 2a shows a path through an alternative embodiment of a network
  • Figure 3 shows a node circuit
  • Figure 3 a shows part of a node circuit
  • Figure 4 illustrates time-slots used during communication
  • Figure 5 illustrates further time-slots used during communication
  • Figure 1 shows a data processing circuit, comprising a plurality of data processing units 10 and a network 12 interconnecting the data processing units.
  • data processing units 10 and network 12 are integrated together in an integrated circuit.
  • the term "data processing unit” is used generically for any type of circuit that produces and/or processes data, including, but not limited to, programmable signal processing cores, dedicated signal processing circuits, such as DCT (Discrete Cosine Transform) computation circuits, memory, IO circuitry etc.
  • DCT Discrete Cosine Transform
  • Terminal node circuits 22a,b couple the data processing units 10a,b to the network 12.
  • pairs of lines are shown for forward and backward communication over communication conductors 20, it should be understood that each line may in fact represent a plurality of communication conductors for message data and various control data for example.
  • part of the communication conductors 20 and node circuits 22 form a communication path through the network between a transmitting data processing unit 10a and a receiving data processing unit 10b. It should be emphasized that the figure merely shows a part of the communication conductors 20 and node circuits 22 in network 12 that is involved with the particular path that is shown.
  • FIG. 2a shows a path through an alternative embodiment of the network, wherein communication conductors 20 are shared for forward and backward communication.
  • an access control circuit 24 which decides which of the node circuits will get access if both node circuits try to get access.
  • selected data processing units 10 transmit streams of data to selected other data processing unit 10 via network 12.
  • Each stream comprises a series of messages that entered network 12 from a transmitting data processing unit 10a and are forwarded from node circuit 22 to node circuit 22 through network 12 to a receiving data processing unit 10b.
  • each message of the stream passes along a path of nodes circuits 22 through network 12.
  • the circuit works in successive time-slots, so that each message is transmitted along successive communication conductors 20 in successive time slots.
  • Data processing units 10 transmit messages from each stream periodically with a network period P, which typically be expressed in terms of a number of transmission time-slots from transmission of one message to transmission of the next message in a stream. All messages from the same stream always pass via the same path.
  • the transmitting data processing unit 10a includes a path description in each message, but the path may also be implied for example by the source and destination of the path.
  • the transmitting data processing unit 10a, or terminal node circuits 22a,b include information in the message to indicate whether the message is an initial message of a stream or not.
  • Figure 3 shows an embodiment of a node circuit 22. This embodiment contains a number of ports 222 coupled to the communication conductors 20 that are coupled to node circuit 22 and respective routing circuits 220 for the respective ports 222.
  • Each routing circuit 220 has an output coupled to a respective outgoing communication conductor 20 from its port and inputs coupled to incoming communication conductors from the remaining ports 20.
  • FIG. 3a shows an embodiment of a part of a routing circuit 220.
  • Routing circuit 220 contains an input message memory 36, reception circuits 37a-c and a forwarding circuit 38.
  • Input message memory 36 is coupled to outgoing data conductor(s) 30d via forwarding circuit 38 and has ports coupled to reception circuits 37a-c.
  • Forwarding circuit 38 is also coupled to an incoming data conductor(s) 32 from the same port as its outgoing data conductor 30d.
  • Each of the reception circuits 37a-c is coupled the incoming data conductors of a respective one of the remaining ports.
  • forwarding circuit 38 has an input for a return enable signal. In operation node circuit 22 receives messages from incoming data conductors
  • Forwarding circuit 38 inspects each message to determine whether the message should be forwarded via the outgoing data conductor(s) 30d that is (are) coupled to the forwarding circuit 38. Forwarding circuit 38 bases the decision to forward a message first of all on whether messages from more than one input 30a-c are present and in selected timeslots also on whether a return message has been received from input 32. In the embodiment wherein two-way or more way communication conductors are used forwarding circuit 38 also makes the decision to forward dependent on whether it can get access to outgoing communication conductor 30d. In this embodiment forwarding circuit issues an access control signal according to the position of the outgoing message in its stream (initial or subsequent).
  • forwarding circuit 38 applies an overruling access signal to the access control circuit and is ensured to be able to transmit the message over its data conductor(s) 30d. If the message is an initial message, forwarding circuit 38 applies a requesting signal to the access control circuit. If the access control circuit 24 that is provided for the communication conductor returns an access grant signal, forwarding circuit 38 transmits the message over its data conductor(s) 30d. If not, the message is not transmitted. In this case the message is lost and said to be discarded. In this embodiment the access control circuit 24 that is provided for a communication conductor returns an access grant signal only if it does not receive an overruling signal from any forwarding circuit (in any attached node circuit).
  • access control circuit 24 receives a single request and no overruling signal it grants the request. If access control circuit 24 receives non-overruling requests from more than one forwarding circuit, it selects one of these forwarding circuits and grants access to that forwarding circuit. Although for the sake of clarity separate routing circuits 220 for each port are shown in figure 3, with memories 36 that store incoming messages from other ports, it will be understood that in practice of course a single routing circuit may be used for all ports, one memory storing received messages from all ports, the routing circuit routing messages to any appropriate port. As will be discussed in the following the source data processing unit 10 that transmits the messages of the stream is informed about the success of transmission of at least the initial message through network 12.
  • the source data processing unit 10 is arranged to refrain from transmitting subsequent messages from the stream if the initial message is discarded anywhere. In this case the source data processing unit 10 will typically try to start the stream anew at a different time point. It should be appreciated that by means of this mechanism, a guaranteed channel is made available, taking up periodically repeating time-slots in node circuits along the path, for the messages of the stream once an initial message of a stream has been successfully forwarded. When an initial message has been forwarded by a node circuit in a time slot it is known that no older streams block access by requiring that node circuit in that time slot. Once transmission of the initial message has succeeded it is ensured that later arising streams cannot prevent access at the expense of subsequent messages. In the following a description will be made of ways of informing a data processing unit 10 about success of transmission of the initial message. In a first embodiment the receiving data processing unit 10b of a stream
  • the original stream (further called “the original stream”) is arranged to set up a return stream in response to reception of the initial message of the original stream.
  • Messages from the return stream are transmitted along the same path along the same node circuits 22 as the messages of the original stream, but in opposite direction.
  • the return stream consists of messages that are transmitted in periodically repeating time-slots.
  • transmitting data processing unit 10a is informed about successful transmission of the initial message of the stream by the arrival of the initial message of the return stream. If transmitting data processing unit 10a does not receive the initial message of the return stream in time, it infers that the initial message has been discarded and ceases transmitting subsequent messages of the original stream.
  • node circuits 22 upon receiving the initial message of the original stream, transmit precursor messages of the return stream back along the path in those periodic time slots wherein subsequently message of the return stream will be transmitted.
  • This enables transmitting data processing unit 10a to detect sooner that the initial message has been discarded, so that it can cease transmitting subsequent messages of the original stream sooner.
  • the node circuits 22 need to know when the messages of the return stream will arrive in future.
  • the circuit is arranged so that this follows from the time of arrival of the initial message of the original stream. In an embodiment, this is facilitated because the receiving data processing unit 10b is arranged to select the time slot of transmission of the initial message of the return stream dependent on the time slot of reception of the initial message of the original stream.
  • This selection is made to ensure that messages from the return stream and the original stream cross each other along the path at specific time points.
  • a crossing is said to occur at a time slot with number t c if a message from the original stream passes on specific data conductor(s) 20 between a pair of node circuits 22 in a time slot with the number -l and a message from the return stream passes between that pair of node circuits 22 in the opposite direction in the time slot with number t c .
  • time-slot of crossing does not depend on the pair of node circuits 22 that is involved, or on the time slot of transmission of the message from the source data processing unit 10.
  • (d) access is available for transmitting a precursor message back along the incoming data conductor(s) 30d in the time slot t
  • the forwarding circuit 38 flags it as an initial message.
  • the precursor message is able to proceed along the path paved by earlier precursor messages, getting overruling access.
  • Figure 4 illustrates occupation of the data conductors along a path as a function of time when this mechanism is used.
  • a transmitting data processing unit 10a transmits an initial message (marked "a") of an original stream in a first time slot to. This initial message travels along successive data conductor(s) along the path in successive time slots. The initial message arrives at the receiving data processing unit 10b in a second time slot t a .
  • receiving data processing unit 10b starts transmitting messages (marked b) of the return stream in a time slot t r , whose slot number is related to t a according to
  • the messages b from the return stream cross the messages a from the original stream (cross, i.e. occupy a data conductor in the time slot after the time slot wherein that data conductor was occupied by a message from the original stream) at crossings 46, 48 at the start of the network periods and halfway the network periods.
  • crossings 40, 42 are simulated.
  • the precursor messages in turn may cause a crossing 44. At such a crossing no new precursor message needs to be generated.
  • the old precursor message p is simply forwarded.
  • the precursor messages p indicate to the transmitting data processing unit 10a of the original stream that the initial message of the original stream has not been discarded along successively further parts of the path through the network.
  • the transmitting data processing unit does not receive a precursor message or normal message of the return stream it does not transmit the next message of the original stream.
  • the precursor messages serve to pave the path for the return stream.
  • the real initial message of the return stream (the first message sent by receiving data processing unit 10b) is in fact a "subsequent" message that follows one or more precursor messages (that have been sent by node circuits 44).
  • the real initial message follows the last preceding precursor message at each node circuit 22 after exactly one network period P. Therefore the real initial message can be treated as a non-initial message that gets unconditional access without the risk of having to cede to other streams.
  • the transmitting data processing unit may have to transmit a subsequent message of the original stream before reception of the first possible precursor message.
  • This in turn has the effect that this subsequent message may attempt to get transmitted along a data conductor where the initial message was discarded because of a collision with another stream.
  • the subsequent message should likewise be discarded. Therefore it is not possible in this case to grant access unconditionally to subsequent messages.
  • the subsequent message preferably contains information representing its sequence number in the original stream.
  • access control circuit 24 grants access to the most senior contending message (having the highest sequence number) and in case of equal seniority a scheme is used that grants consistent access to the same stream.
  • sequence number generalizes the distinction between initial and subsequent messages.
  • the sequence number need not represent sequence numbers higher than the maximum number of messages that can be transmitted with a network period P before the transmitting data processing unit will receive confirmation of arrival of the initial message.
  • the maximum sequence number may be used in all messages with higher sequence number.
  • the node circuits send precursor tokens in the time slots that satisfy the expression for t c .
  • the offset value "s" may be chosen to be equal for all streams, or it may be selected independently for each stream. In the latter case, the offset value may be specified in the messages (or at least the leading messages of each stream), and the node circuits may be arranged to select time slots of sending the precursor tokens according to the expression for t c as a function of the specified offset.
  • the specification of the offset value may be explicit or inherent, e.g. as a result of a type indication of the stream, or dependent on the source or destination of the stream. If the length "k" of the path is known in advance (and can be derived from the messages), the offset value "s" can even be specified by specifying a time slot numbers t r of sending return messages. However, it will be appreciated that the embodiment that has been described has the advantage that the time of crossing can be determined without any advance knowledge about the length of the time path. As a result, the path can be selected dynamically. It must be appreciated that in other embodiments no return stream needs to be involved. In this case, what have been called precursor messages in the preceding merely serve as confirmation messages of forwarding.
  • the messages may specify the path, so that the nodes will route the messages dependent on the specified path.
  • the node circuits may be arranged to route the messages along a reproducible path, for example by retaining information about the selected route for a previous message from the same stream, or by using a predetermined route from a given source to a given destination.
  • the path specification may be selected in advance by the transmitting data processing unit 10a, or it may be derived from a path selected by the node circuits for the initial message of the stream.
  • information about the path may be added to the initial message as it travels along the path, and the node circuit may report back this information in the precursor messages, so that the transmitting data processing unit 10a can include a specification in subsequent messages.
  • the node circuits may cross-transfer path information between the crossing messages of the original stream and the return stream.
  • the messages may each time receive information about the path to be followed to the next crossing.
  • the node circuit may transfer information that specifies the path followed by a message from the original stream from the message of the return stream. In this way information about the path to the time of next crossing (or further) may be provided to the message from the original stream.
  • the path can be dynamically selected at the time of forwarding of the initial message and no path information needs to be kept in the nodes.
  • similar information may be transferred from the message of the original stream to the message of the return stream. This saves on the amount of data needed to represent the path, but is not strictly necessary, since messages of the return stream can be provided with information about the backward path ab initio.
  • the node circuits may be arranged to buffer the messages for a number of time-slots, for example to wait for the availability of a data conductor.
  • condition (b) for sending a precursor message is generalized to (b') a message from the original stream has arrived at the node circuit form the incoming data conductor(s) 32 in the preceding time slot t-1, or this message has arrived in an earlier time- slot and is still buffered awaiting transmission.
  • Condition (d) for sending a precursor message is relaxed, because the precursor message may remain buffered for a number of time slots before transmission. Instead, the precursor message is preferably discarded if transmission is delayed so much that the precursor message does not, or cannot, arrive in before the time-slot for the next crossing time slot at the node circuit where that crossing has to take place (as determined by the transmission of an earlier precursor from that node circuit).
  • Figure 5 illustrates an example of occupation of the data conductors along a path as a function of time when this mechanism is used.
  • the initial message of the original stream is buffered in a time slot.
  • the new precursor message corresponding to this initial message is transmitted in the normal time-slot as described before, but from another node circuit than in the absence of delay.
  • the new precursor message crosses a message from the original stream in a time slot 52 at a node circuit a whole network period P after a time slot 40 in which an older precursor was transmitted from that node circuit.
  • the precursor must be delayed, but this delay may occur anywhere along the path.
  • This condition may be relaxed by introducing an intentional delay each time before transmitting a precursor message after the time-slot for which it is generated. Thus, the time constraint on the arrival of precursor messages from node circuit further down the path is relaxed.
  • the intentional delay may be set dependent on the delay suffered by the initial message of the original stream, increasing the intentional delay of the precursor message if the initial message has suffered little delay, and decreasing the delay of the precursor message if the initial message has suffered more delay.
  • each communication conductor 20 may support transmission of a plurality of messages per time slot. In an alternative embodiment this may be used to transmit messages from multiple streams of periodic messages in the same time slot and/or individual messages that do not belong to a periodic stream.
  • the number of messages that may be transmitted in a time slot along a communication conductor 20 may be a predetermined number of messages, or the number of messages may depend on the combined sizes of the messages. In the latter case the messages of a given stream preferably all have the same size, or should at least claim the same (maximum) size.
  • the node circuits preferably use the claimed size of a given stream to limit the number of other streams of periodic messages for transmission in the time slot along the same communication conductor. This number of other streams is limited so that their combined claimed size leaves sufficient capacity the time slot for the maximum size of the messages from the given stream. If a message requires from the given stream less than the claimed size (or even no size at all), the node circuit may use the remaining bandwidth may be used for transmission of individual messages.
  • each message may extend over a plurality of time slots. In this case transmission basically comes down to transmission of multiple streams in successive time slots.
  • the invention has been described for handling bandwidth limitations on communication conductors 20, it will be appreciated that the invention may be applied to handle other types of resource limitations in the network as well.
  • the invention has been illustrated by means of a particular circuit, other circuit implementations are possible. For example, instead of a structured node circuit 22 as shown in figures 3 and 3a a simplified node circuit may be used with a processing circuit that processes inputs messages from all communication conductors 20 and transmits messages at selected communication conductors 20.
  • data processing units 10 have been shown that are connected to a communication conductor 20 via terminal node circuits 22a,b, one may of course include the function of the terminal node circuits 22a,b in the data processing units 10a,b (in which the data processing unit is considered to be a node circuit as well), or in multi-input node circuits.
  • terminal node circuits 22a,b the data processing circuits 10 have access to the terminal node circuits without contention, the terminal node circuits 22a,b handling access to the communication conductors 20.

Abstract

L'invention concerne des données traitées dans un circuit qui contient plusieurs unités de traitement de données interconnectées par un réseau de circuits de noeud. Les circuits de noeud utilisent des ressources sur la base du multiplexage d'intervalles de temps. Un flux est initié entre des unités de traitement de données sources et des unités de traitement de données de destination. Chaque flux contient des messages qui occupent les ressources dans une sélection périodiquement répétée d'intervalles de temps. La période de répétition consiste en une période de réseau qui est identique pour tous les flux. Les circuits de noeud transmettent tous les messages d'un flux particulier à travers le réseau le long des noeuds dans un trajet spécifique de flux attribué au flux particulier. Les circuits de noeud décident de la transmission ou non de chaque message en fonction d'une mesure d'ancienneté du message dans son flux particulier. Un circuit de noeud empêche la transmission d'un message moins ancien pour lequel des ressources insuffisantes sont laissées par transmission de messages plus anciens à partir du circuit de noeud particulier. De préférence, dans des intervalles de temps de retour sélectionnés, les circuits de noeud envoient des confirmations afin de confirmer la transmission avec succès d'un message. Les intervalles de temps de retour sélectionnés sont sélectionnés de façon prédéfinie de sorte qu'il est possible de prédire dans un circuit de noeud situé plus loin sur le trajet, de la présence d'un autre message du flux au niveau de ce circuit de noeud dans un intervalle de temps de croisement, si le message croise la confirmation au niveau du circuit de noeud ultérieur (22) dans l'intervalle de temps de coincement, par exemple par envoi de confirmations à partir de ces circuits de noeud (22) dans lesquels le message initial est arrivé dans une des séries d'intervalles de temps qui se répète chaque demi-période de réseau. Les circuits de noeud détectent si un message du flux dans les circuits de noeud croisent avec la confirmation les intervalles de temps de croisement attendus. Les circuits de noeud transmettent des messages uniquement si le croisement est détecté dans des intervalles de temps attendus. De préférence les confirmations sont transmises de sorte que des confirmations successives occupent les ressources périodiquement avec la période de réseau. Dans ce cas, l'unité de traitement de données de destination peut débuter un flux de messages de retour, dans des intervalles de temps qui suivent les confirmations après la période de réseau. La transmission de ce flux de retour est assurée à l'avance.
PCT/IB2005/050626 2004-03-03 2005-02-21 Circuit de traitement de donnees dans lequel des unites de traitement de donnees communiquent via un reseau WO2005088464A2 (fr)

Priority Applications (5)

Application Number Priority Date Filing Date Title
KR1020067017642A KR101125639B1 (ko) 2004-03-03 2005-02-21 네트워크를 통하여 통신이 가능한 데이터 처리 유닛을 가지는 데이터 처리 회로
US10/598,299 US7839774B2 (en) 2004-03-03 2005-02-21 Data processing circuit wherein data processing units communicate via a network
EP05708805A EP1726132A2 (fr) 2004-03-03 2005-02-21 Circuit de traitement de donnees dans lequel des unites de traitement de donnees communiquent via un reseau
CN2005800066830A CN1926826B (zh) 2004-03-03 2005-02-21 其中数据处理单元经网络进行通信的数据处理电路
JP2007501388A JP4598051B2 (ja) 2004-03-03 2005-02-21 データ処理ユニットがネットワークを介し通信するデータ処理回路

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
EP04100849.1 2004-03-03
EP04100849 2004-03-03

Publications (2)

Publication Number Publication Date
WO2005088464A2 true WO2005088464A2 (fr) 2005-09-22
WO2005088464A3 WO2005088464A3 (fr) 2005-12-01

Family

ID=34960718

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/IB2005/050626 WO2005088464A2 (fr) 2004-03-03 2005-02-21 Circuit de traitement de donnees dans lequel des unites de traitement de donnees communiquent via un reseau

Country Status (6)

Country Link
US (1) US7839774B2 (fr)
EP (1) EP1726132A2 (fr)
JP (1) JP4598051B2 (fr)
KR (1) KR101125639B1 (fr)
CN (1) CN1926826B (fr)
WO (1) WO2005088464A2 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010105425A1 (fr) * 2009-03-19 2010-09-23 上海贝尔股份有限公司 Procédé, dispositif et programme d'ordinateur pour interaction d'informations partagées entre des unités de communication
WO2014146302A1 (fr) * 2013-03-22 2014-09-25 华为技术有限公司 Procédé et appareil de transmission de données de liaison montante

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020126649A1 (en) * 1986-09-16 2002-09-12 Yoshito Sakurai Distributed type switching system

Family Cites Families (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE3305130A1 (de) * 1983-02-15 1984-08-16 Kuka Umwelttechnik GmbH, 8900 Augsburg Muellsammelfahrzeug
US4731878A (en) * 1985-11-29 1988-03-15 American Telephone And Telegraph Company, At&T Bell Laboratories Self-routing switch node combining electronic and photonic switching
US4785446A (en) * 1986-11-07 1988-11-15 International Business Machines Corporation Distributed bit switching of a multistage interconnection network
JPH04291548A (ja) * 1991-03-20 1992-10-15 Fujitsu Ltd 高速大容量マトリクス型時間分割ラベル交換方式
JPH0818590A (ja) * 1994-06-30 1996-01-19 Toshiba Corp ループ式通信システム
JP2607851B2 (ja) * 1994-09-17 1997-05-07 株式会社東芝 プロセッサ間ルーティング方式
SE513221C2 (sv) * 1998-04-17 2000-07-31 Net Insight Ab Förfarande och anordning för allokering av tidluckor till en kanal i ett kretskopplat tidsmultiplexerat nät
US7206280B1 (en) * 2000-09-12 2007-04-17 Lucent Technologies Inc. Method and apparatus for asynchronous incremental redundancy reception in a communication system
JP2002281559A (ja) * 2001-03-19 2002-09-27 Nec Eng Ltd 移動通信における発信予約システム
US7085266B2 (en) * 2001-03-21 2006-08-01 International Business Machines Corporation Apparatus, method and limited set of messages to transmit data between components of a network processor
US7042848B2 (en) * 2001-05-04 2006-05-09 Slt Logic Llc System and method for hierarchical policing of flows and subflows of a data stream
JP3473951B2 (ja) * 2001-06-01 2003-12-08 Ｎｅｃマイクロシステム株式会社 データ処理装置およびシステム
US7295624B2 (en) * 2002-03-06 2007-11-13 Texas Instruments Incorporated Wireless system with hybrid automatic retransmission request in interference-limited communications
US7352769B2 (en) * 2002-09-12 2008-04-01 Intel Corporation Multiple calendar schedule reservation structure and method
US7769893B2 (en) * 2002-10-08 2010-08-03 Koninklijke Philips Electronics N.V. Integrated circuit and method for establishing transactions
US7640446B1 (en) * 2003-09-29 2009-12-29 Marvell International Ltd. System-on-chip power reduction through dynamic clock frequency

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020126649A1 (en) * 1986-09-16 2002-09-12 Yoshito Sakurai Distributed type switching system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
DIELISSEN J ET AL: "Concepts and Implementation of the Philips Network-on-Chip" INTERNET ARTICLE, 13 November 2003 (2003-11-13), pages 1-6, XP002330547 Eindhoven *
MITTAL A ET AL: "Dynamic real-time channel establishment in multiple access bus networks" COMPUTER COMMUNICATIONS, ELSEVIER SCIENCE PUBLISHERS BV, AMSTERDAM, NL, vol. 26, no. 2, 1 February 2003 (2003-02-01), pages 113-127, XP004401228 ISSN: 0140-3664 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010105425A1 (fr) * 2009-03-19 2010-09-23 上海贝尔股份有限公司 Procédé, dispositif et programme d'ordinateur pour interaction d'informations partagées entre des unités de communication
CN102273171A (zh) * 2009-03-19 2011-12-07 上海贝尔股份有限公司 一种用于在通信单元之间交互共享信息的方法、设备和计算机程序
WO2014146302A1 (fr) * 2013-03-22 2014-09-25 华为技术有限公司 Procédé et appareil de transmission de données de liaison montante
CN104488244A (zh) * 2013-03-22 2015-04-01 华为技术有限公司 上行数据传输方法及装置
CN104488244B (zh) * 2013-03-22 2017-06-27 华为技术有限公司 上行数据传输方法及装置
US9985770B2 (en) 2013-03-22 2018-05-29 Huawei Technologies Co., Ltd. Uplink data transmission method and apparatus

Also Published As

Publication number Publication date
EP1726132A2 (fr) 2006-11-29
KR20060130192A (ko) 2006-12-18
JP4598051B2 (ja) 2010-12-15
JP2007526710A (ja) 2007-09-13
WO2005088464A3 (fr) 2005-12-01
US7839774B2 (en) 2010-11-23
CN1926826A (zh) 2007-03-07
US20080259949A1 (en) 2008-10-23
KR101125639B1 (ko) 2012-03-27
CN1926826B (zh) 2010-10-20

Similar Documents

Publication Publication Date Title
US5675573A (en) Delay-minimizing system with guaranteed bandwidth delivery for real-time traffic
US7352765B2 (en) Packet switching fabric having a segmented ring with token based resource control protocol and output queuing control
JP4995101B2 (ja) 共有リソースへのアクセスを制御する方法及びシステム
US6246692B1 (en) Packet switching fabric using the segmented ring with resource reservation control
US4500987A (en) Loop transmission system
EP0179550B1 (fr) Réseau étoile commandé
US4292623A (en) Port logic for a communication bus system
US5898691A (en) Method and apparatus for congestion distributed adaptive routing
JPH10190710A (ja) アクセス調停方法
EP0505658B1 (fr) Technique d'accès au support de transmission de réseaux locaux
KR20040054721A (ko) 컴퓨터 시스템의 입력/출력 노드에서 태깅 및 중재 매카니즘
JPH01132246A (ja) トラヒックの片寄りを無くすことができるデータ伝送システム
KR100968250B1 (ko) 컴퓨터 시스템 ｉ/ｏ 노드
US10389806B2 (en) Method for asynchronous data communication in a real-time capable ethernet data network
Kranich et al. NoC switch with credit based guaranteed service support qualified for GALS systems
EP2323317A1 (fr) Procédé de commande de bande et dispositif de commande de bande pour dispositif de n ud
US7839774B2 (en) Data processing circuit wherein data processing units communicate via a network
US6643702B1 (en) Traffic scheduler for a first tier switch of a two tier switch
CN101129033B (zh) 控制对共享资源的访问的方法和系统
JPH04328929A (ja) Ｌａｎシステムにおけるシグナリング情報の伝送方法及び装置
US7345995B2 (en) Conflict resolution in data stream distribution
JPH1084348A (ja) 非同期転送モード通信方式におけるセル送信装置およびセル送信制御方法
JPWO2002035774A1 (ja) データ多重化ネットワーク機構
Yotsui et al. A high‐throughput flow control scheme using a relay buffering function
Chan Ring Networks

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SM SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
WWE Wipo information: entry into national phase

Ref document number: 2005708805

Country of ref document: EP

WWE Wipo information: entry into national phase

Ref document number: 1020067017642

Country of ref document: KR

WWE Wipo information: entry into national phase

Ref document number: 200580006683.0

Country of ref document: CN

Ref document number: 2007501388

Country of ref document: JP

NENP Non-entry into the national phase

Ref country code: DE

WWW Wipo information: withdrawn in national office

Country of ref document: DE

WWP Wipo information: published in national office

Ref document number: 2005708805

Country of ref document: EP

WWP Wipo information: published in national office

Ref document number: 1020067017642

Country of ref document: KR

WWE Wipo information: entry into national phase

Ref document number: 10598299

Country of ref document: US